So far, there have been plenty of literatures on the metric in the space of probability distributions and quantum states. As for channels, however, only a little had been known. In this paper, we impose monotonicity by concatenation of channels before and after the given channel families, and invariance by tensoring identity channels. Under these axioms, we identify the largest and the smallest metrics. Also, we studied asymptotic theory of metric in parallel and adaptive repetition settings, and applied them to the study of channel estimation. First we express the achievable lower bound of the mean square error (MSE) of an estimate by a monotone channel metric, and show this equals O(1/n) for noisy channels, where n is the number of times of channel use. This result shows Heisenberg rate, or O(1/n^2)-rate of the MSE observed in case of estimation of unitary, collapses with very small arbitrary noise.